Remove redundant work declaration in HISI_SAS_DECLARE_RST_WORK_ON_STACK

Link: https://lore.kernel.org/r/1567774537-20003-10-git-send-email-john.garry@huawei.comSigned-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

We never call hisi_sas_hw.slot_complete, so remove it.

Link: https://lore.kernel.org/r/1567774537-20003-9-git-send-email-john.garry@huawei.comSigned-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Currently the NCQ tag is only assigned for FPDMA READ and FPDMA WRITE
commands, and for other NCQ commands (such as FPDMA SEND), their NCQ tags
are set in the delivery command to 0.

So for all the NCQ commands, we also need to assign normal NCQ tag for
them, so drop the command type check in hisi_sas_get_ncq_tag() [drop
hisi_sas_get_ncq_tag() altogether actually], and always use the ATA command
NCQ tag when appropriate.

Link: https://lore.kernel.org/r/1567774537-20003-8-git-send-email-john.garry@huawei.comSigned-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The AXI and RAS register values should also should be snapshot at debugfs.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The value of IOST/ITCT is updated to cache first, and then synchronize to
DDR periodically. So the value in IOST/ITCT cache is the latest data and
it's important for debugging.

So, the HW cache of IOST and ITCT should be snapshot at debugfs.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

In commit 1273d65f ("scsi: hisi_sas: change queue depth from 512 to
4096"), the depth of each queue is the same as the max IPTT in the system.

As such, as long as we have an IPTT allocated, we will have enough space on
any delivery queue.

All .get_free_slot functions were checking for space on the queue by
reading the DQ read pointer. Drop this, and also raise the code into common
code, as there is nothing hw specific remaining.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

There is a small optimisation to be had by making the max IPTT the same for
all hw revisions, that being we can drop the check for read and write
pointer being the same in the get free slot function.

Change v1 hw to have max IPTT of 4096 - same as v2 and v3 hw - and
drop hisi_sas_hw.max_command_entries.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Many times in libsas, and in LLDDs which use libsas, the check for an
expander device is re-implemented or open coded.

Use dev_is_expander() instead. We rename this from
sas_dev_type_is_expander() to not spill so many lines in referencing.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Reviewed-by: NJason Yan <yanaijie@huawei.com>
Reviewed-by: NJack Wang <jinpu.wang@cloud.ionos.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Macro HISI_SAS_SGE_PAGE_CNT is defined to SG_CHUNK_SIZE, which is 128.

This means that sizeof(struct hisi_sas_slot_buf_table) is 4192. This is
just over a 4K, which can mean inefficient DMA memory usage (for no PI).

Reduce the size of HISI_SAS_SGE_PAGE_CNT to 124 to fit in a 4K page. With
this change, we experience no performance hit.

Cc: dann frazier <dann.frazier@canonical.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Based on 1 normalized pattern(s):

  this program is free software you can redistribute it and or modify
  it under the terms of the gnu general public license as published by
  the free software foundation either version 2 of the license or at
  your option any later version

extracted by the scancode license scanner the SPDX license identifier

  GPL-2.0-or-later

has been chosen to replace the boilerplate/reference in 3029 file(s).
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NAllison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070032.746973796@linutronix.deSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

In commit efdcad62 ("scsi: hisi_sas: Set PHY linkrate when
disconnected"), we use the sas_phy_data.enable flag to track whether the
PHY was enabled or not, so that we know if we should set the PHY negotiated
linkrate at SAS_LINK_RATE_UNKNOWN or SAS_PHY_DISABLED.

However, it is not proper to use sas_phy_data.enable, since it is only set
when libsas attempts to set the PHY disabled/enabled; hence, it may not
even have an initial value.

As a solution to this problem, introduce hisi_sas_phy.enable to track
whether the PHY is enabled or not, so that we can set the negotiated
linkrate properly when the PHY comes down.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If we exchange SAS expander from one SAS controller to other SAS controller
without powering it down, the STP target port will maintain previous
affiliation and reject all subsequent connection requests from other STP
initiator ports with OPEN_REJECT (STP RESOURCES BUSY).

To solve this issue, send HARD RESET to clear the previous affiliation of
STP target port according to SPL (chapter 6.19.4).

We (re-)introduce dev status flag to know if to sleep in NEXUS reset code
or not for remote PHYs. The idea is that if the device is being
initialised, we don't require the delay, and caller would wait for link to
be established, cf. sas_ata_hard_reset().
Co-developed-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The later revision of v3 hw has added an function of interrupt coalesce
according to time for PHY RX errors. We set the coalesce time to 1s.  Then
we print PHY RX errors count when PHY RX errors happen, and don't need to
worry that there may be too much log prints.

Besides, we use hisi_sas_phy.lock to protect error count value. Because we
update them by calling phy_get_events_v3_hw(), which is also used by core
driver (for get PHY events function).

We relocate phy_get_events_v3_hw() to avoid a further declaration.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

For auto-control irq affinity mode, choose the dq to deliver IO according
to the current CPU.

Then it decreases the performance regression that fio and CQ interrupts are
processed on different node.

For user control irq affinity mode, keep it as before.

To realize it, also need to distinguish the usage of dq lock and sas_dev
lock.

We mark as experimental due to ongoing discussion on managed MSI IRQ
during hotplug:
https://marc.info/?l=linux-scsi&m=154876335707751&w=2

We're almost at the point where we can expose multiple queues to the upper
layer for SCSI MQ, but we need to sort out the per-HBA tags performance
issue.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

To support queue mapped to a CPU, it needs to be ensured that issuing an
internal abort is safe, in that it is guaranteed that an internal abort is
processed for a single IO or a device after all the relevant command(s)
which it is attempting to abort have been processed by the controller.

Currently we only deliver commands for any device on a single queue to
solve this problem, as we know that commands issued on the same queue will
be processed in order, and we will not have a scenario where the internal
abort is racing against a command(s) which it is trying to abort.

To enqueue commands on queue mapped to a CPU, choosing a queue for an
command is based on the associated queue for the current CPU, so this is
not safe for internal abort since it would definitely not be guaranteed
that commands for the command devices are issued on the same queue.

To solve this issue, we take a bludgeoning approach, and issue a separate
internal abort on any queue(s) relevant to the command or device, in that
we will be guaranteed that at least one of these internal aborts will be
received last in the controller.

So, for aborting a single command, we can just force the internal abort to
be issued on the same queue as the command which we are trying to abort.

For aborting all commands associated with a device, we issue a separate
internal abort on all relevant queues. Issuing multiple internal aborts in
this fashion would have not side affect.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If sending IOs to many disks from single queue, it is possible that the
queue may be full. To avoid the situation, change queue depth from 512 to
4096 which is the max number of IOs for v3 hw.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch adds support for DIX to v3 hw driver.

For this, we build upon support for DIF, most significantly is adding new
DMA map and unmap paths.

Some pre-existing macro precedence issues are also tidied. They were
detected by checkpatch --strict.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Hot-plugging SAS wire of direct hard disk backplane may cause disk lost. We
have done this test with several types of SATA disk from different venders,
and only two models from Seagate has this problem, ST4000NM0035-1V4107 and
ST3000VM002-1ET166.

The root cause is that the disk doesn't send D2H frame after OOB finished.
SAS controller will issue phyup interrupt only when D2H frame is received,
otherwise, will be waiting there all the time.

When this issue happen, we can find the disk again with link reset.  To fix
this issue, we setup an timer after OOB finished. If the PHY is not up in
20s, do link reset. Notes: the 20s is an experience value.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

In function hisi_sas_alloc(), parameter shost is not used, so remove it.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When issing a hardreset to a SATA device when running IO, it is possible
that abnormal CQs of the device are returned. Then enter error handler, it
doesn't enter function hisi_sas_abort_task() as there is no timeout IO, and
it doesn't set device as HISI_SAS_DEV_EH. So when hardreset by libata
later, it actually doesn't issue hardreset as there is a check to judge
whether device is in error.

For this situation, actually need to hardreset the device to recover.
So remove the check of sas_dev status in hisi_sas_I_T_nexus_reset().

Before we add the check to avoid the endless loop of reset for
directly-attached SATA device at probe time, actually we flutter it for
it, so it is not necessary to add the check now.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Send primitive NOTIFY to SSP situation only, or it causes underflow issue
when sending IO. Also rename hisi_sas_hw.sl_notify() to hisi_sas_hw.
sl_notify_ssp().
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Sparse can detect some type casting issues in the debugfs code, so fix it
up.

Also a missing static qualifier is added to hisi_sas_debugfs_to_reg_name().
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When calling debugfs functions, there is no need to ever check the return
value. The function can work or not, but the code logic should never do
something different based on this.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch create debugfs file for global register and add file
operations.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch takes snapshot for global regs, port regs, CQ, DQ, IOST, ITCT.

Add code for snapshot trig and generate dump directory.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch allocates snapshot memory for global reg, port regs, CQ, DQ,
IOST, ITCT.

When we fail to allocate memory for some registers, we free the memory
and set hisi_sas_debugfs_enable as 0 to stop loading debugfs from running.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch creates root directory at hisi_sas_init() and generates device
directory when we probe device driver.

And we remove the root directory at hisi_sas_exit(), but recursively
delete device directory when we remove device driver.
Signed-off-by: NLuo Jiaxing <luojiaxing@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

For v3 hw, we support DIF operation for SAS, but not SATA.

In addition, DIF CRC16 is supported.

This patchset adds the SW support for the described features. The main
components are as follows:

- Get protection mask from module param
- Fill PI fields
- Fill related to DIF in DQ and protection iu memories
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patchset fixes some warnings detected by the sparse tool, like these:
drivers/scsi/hisi_sas/hisi_sas_main.c:1469:52: warning: incorrect type in assignment (different base types)
drivers/scsi/hisi_sas/hisi_sas_main.c:1469:52:    expected unsigned short [unsigned] [assigned] [usertype] tag_of_task_to_be_managed
drivers/scsi/hisi_sas/hisi_sas_main.c:1469:52:    got restricted __le16 [usertype] <noident>
drivers/scsi/hisi_sas/hisi_sas_main.c:1723:52: warning: incorrect type in assignment (different base types)
drivers/scsi/hisi_sas/hisi_sas_main.c:1723:52:    expected unsigned short [unsigned] [assigned] [usertype] tag_of_task_to_be_managed
drivers/scsi/hisi_sas/hisi_sas_main.c:1723:52:    got restricted __le16 [usertype] <noident>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

If INT_COAL_EN is enabled, configure time and count of interrupt
coalescing.  Then if CQ collects count of CQ entries in time, it will
report the interrupt. Or if CQ doesn't collect enough CQ entries in time,
it will report the interrupt at timeout.

As all the registers are not supported to be changed dynamically, we need
to config those register between disable and enable PHYs.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Currently all the three HBA (v1/v2/v3 HW) share the same host attributes.

To support each HBA having separate attributes in future, create per-HBA
attributes.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Currently we use the IPTT defined in LLDD to identify IOs. Actually for
IOs which are from the block layer, they have tags to identify them. So
for those IOs, use tag of the block layer directly, and for IOs which is
not from the block layer (such as internal IOs from libsas/LLDD), reserve
96 IPTTs for them.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch tidies host controller reset function by putting some code to
two new functions, and exports these two functions out, so that they could
be used by FLR feature to be realised.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

For some time now we have not used hisi_sas_slot_abort() to handle erroring
slots, apart from in archaic v1 hw.

As such, remove this function and associated code. For v1 hw, move error
handling to same scheme as other hw revisions, where we allow erroring
commands to timeout.
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Currently the driver spends much time allocating and freeing the slot DMA
buffer for command delivery/completion. To boost the performance,
pre-allocate the buffers for all IPTT. The downside of this approach is
that we are reallocating all buffer memory upfront, so hog memory which we
may not need.

However, the current method - DMA buffer pool - also caches all buffers and
does not free them until the pool is destroyed, so is not exactly efficient
either.

On top of this, since the slot DMA buffer is slightly bigger than a 4K
page, we need to allocate 2x4K pages per buffer (for 4K page kernel), which
is quite wasteful. For 64K page size this is not such an issue.

So, for the 4K page case, in order to make memory usage more efficient,
pre-allocating larger blocks of DMA memory for the buffers can be more
efficient.

To make DMA memory usage most efficient, we would choose a single
contiguous DMA memory block, but this could use up all the DMA memory in
the system (when CMA enabled and no IOMMU), or we may just not be able to
allocate a DMA buffer large enough when no CMA or IOMMU.

To decide the block size we use the LCM (least common multiple) of the
buffer size and the page size. We roundup(64) to ensure the LCM is not too
large, even though a little memory may be wasted per block.

So, with this, the total memory requirement is about is about 17MB for 4096
max IPTT.

Previously (for 4K pages case), it would be 32MB (for all slots
allocated).

With this change, the relative increase of IOPS for bs=4K read when
PAGE_SIZE=4K and PAGE_SIZE=64K is as follows:
    IODEPTH     4K PAGE_SIZE      64K PAGE_SIZE
    32          56%               47%
    64          53%               44%
    128         64%               43%
    256         67%               45%
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

There is a possible conflict when a device is removed and host reset occurs
concurrently.

The reason is that then the device is notified as gone, we try to clear the
ITCT, which is notified via an interrupt. The dev gone function pends on
this event with a completion, which is completed when the ITCT interrupt
occurs.

But host reset will disable all interrupts, the wait_for_completion() may
wait indefinitely.

This patch adds an semaphore to synchronise this two processes. The
semaphore is taken by the host reset as the basis of synchronising.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

When issuing a nexus reset for directly attached device, we want to ignore
the PHY down events so libsas will not deform and reform the port.

In the case that the attached SAS changes for the reset, libsas will deform
and form a port.

For scenario that the PHY does not come up after a timeout period, then
report the PHY down to libsas.
Signed-off-by: NXiang Chen <chenxiang66@hisilicon.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

This patch adds a force PHY function for internal ATA command for v2 hw.

Because there is an SoC bug in v2 hw, and need send an IO through each PHY
of a port to work around a bug which occurs after a controller reset.

This force PHY function will be used in the later patch.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

In future scenarios we will want to use the TMF struct for more task types
than SSP.

As such, we can add struct hisi_sas_tmf_task directly into struct
hisi_sas_slot, and this will mean we can remove the TMF parameters from the
task prep functions.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

We may reset the controller in many scenarios, such as SCSI EH and HW
errors. There should be no IO which returns from target when SCSI EH is
active. But for other scenarios, there may be.  It is not necessary to make
such IOs fail.

This patch adds an function of trying to wait for any commands, or IO, to
complete before host reset. If no more CQ returned from host controller in
100ms, we assume no more IO can return, and then stop waiting. We wait 5s
at most.

The HW has a register CQE_SEND_CNT to indicate the total number of CQs that
has been reported to driver. We can use this register and it is reliable to
resd this register in such scenarios that require host reset.
Signed-off-by: NXiaofei Tan <tanxiaofei@huawei.com>
Signed-off-by: NJohn Garry <john.garry@huawei.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>